AWS · 703

Anthropic 40% of ARR is sold by cloud vendors: $65 billion in annualized revenue is not that easy to earn

Comparative news, AI news, Anthropic's annualized revenue just reached $65 billion, and SemiAnalysis then split this revenue. According to its model estimates, more than 40% of ARR in the second quarter came from indirect channels such as AWS Bedrock, Microsoft Foundry, and Google's enterprise AI platform. What is worth paying attention to is how much profit these revenues can leave behind. In Bedrock, for example, Claude was sold by Anthropic. Anthropic will count the total amount of tokens sold into ARR, and then pay AWS for computing power and channel sharing. In other words, it's also a $1 ARR. If you sell it through a cloud platform, Anthropic will end up leaving less money than direct sales. So while $65 billion ARR isn't fake income, the revenue structure is clearly not that healthy. The higher the share of channels, the less direct equations between revenue growth and profit growth. If you only look at ARR, you might be overestimating the contribution of these revenues to Anthropic's final profit. Of course, there are benefits to the channel model. AWS, Microsoft, and Google already have a large number of enterprise customers and procurement contracts, and can directly cram Claude into existing cloud bills. Anthropic is now trading some of its profits for scale and customer acquisition efficiency.

3d ago

Fireworks that came out of Meta to talk about open source and closed source. Who will win?

Author: Silicon Valley Vector Silicon Valley Coordinates Editor: Peggy, BlockBeats Original title: Silicon Valley Coordinates x Fireworks Co-Founder Benny Chen: Open Source Models, Token Growth, Inference Optimization, and Model Customization Editor's Note: In the context of open source models speeding up and approaching cutting-edge closed-source models, industry discussions are shifting from “who has the most capable model” to “who can put models into production at a lower cost”. However, when model capabilities converged and token consumption increased, a lower-level question began to emerge: are companies really willing to pay a cheaper model call, or exclusive intelligence that can perform specific tasks in a stable manner? Recently, Cao Qingyun, host of “Silicon Valley Coordinates”, had a conversation with Chen Yufei, co-founder of Fireworks AI. Located between models and enterprise applications, Fireworks mainly provides customers with open source model inference, performance optimization, and customization services. Rather than simply discussing whether open source can catch up with closed sources, Chen Yufei's observations are closer to actual workloads: where tokens flow, why companies pay, and what is still missing from the model from proof of concept to production. In this conversation, Chen Yufei disassembled “who wins between open source and closed source” into a set of lower level structural questions: can token growth be converted into revenue, can generic capabilities replace vertical accumulation, can the low price model pass corporate evaluation, and how the inference platform can gain value between cloud vendors and application companies. First, the scale of use and commercial value of the open source model are diverging. In the past, the ability to catch up and call price were the main indicators for judging the competitiveness of open source; today, the Fireworks platform processes about 40 trillion to 50 trillion tokens every day, and the actual usage of the open source model has rapidly expanded. However, free traffic, promotional subsidies, and model price differences will cause Token statistics to overestimate some demand. Customers may heavily use lower-cost models and still hand over the highest budget to the best-performing closed source model. This means that the next phase of open source is not just expanding traffic, but proving that it can meet or even surpass cutting-edge models for high-value tasks, and turn cost advantages into willingness to pay. Second, the general model and the vertical model are beginning to evolve in different directions. In the past, every time a cutting-edge model was upgraded, it was possible to directly eliminate a number of fine-tuned models; now, vertical applications such as law, medical care, and programming are accumulating more detailed evaluations, data, and workflows, and their optimization goals are gradually separated from cutting-edge laboratories. Generic models need to increase the upper limit of capabilities, while vertical models require stable delivery of results in limited scenarios. The former can solve a wider range of problems, while the latter has a better understanding of how users define “right.” This means that the barrier for vertical companies is not just having a customized model, but being able to continuously transform industry needs into an evaluation system and migrate over and over again as the basic model is updated. Third, the bottleneck in enterprise AI implementation is shifting from model supply to evaluation capabilities. In the past, enterprise proof of concept often relied on trial experience and subjective judgment; now, when AI enters production processes such as call centers, legal searches, and medical assistance, it is no longer possible to support procurement decisions simply by “looking good.” Businesses must know what tasks the model works for, when it fails, and how much the cost and quality of switching from closed source to open source changes. Assessment is therefore no longer an ancillary tool, but an infrastructure connecting procurement, training, and production deployment. Who can define tasks, establish test distributions, and continuously update standards can truly control model choices. Fourth, the value of inference platforms is shifting from “selling cheap computing power” to organizational models, hardware, and workflows. In the past, inference optimization was mainly understood to reduce the cost of a single token; now, caching, task splitting, model routing, and context management can all directly change the task completion rate. Different models don't have to compete for the same position; they can act as performers and advisors separately. Fireworks' business logic is also based on this: instead of building asset-heavy hardware, revenue is tied to actual use of customer models through training, customization, and continuous reasoning. But the main rival in this path is not a single new cloud company, but a large cloud vendor that can simultaneously control computing power, software, and customer portals. Fifth, the rise of the open source model may not reduce infrastructure requirements; on the contrary, it may reduce model layer premiums and further push value towards reasoning and computing power. Tech giants continue to increase capital spending, not just calculating short-term returns, but measuring missed AI cycles...

4d ago律动BlockBeats#AI

Anthropic ARR sparks valuation controversy: behind $65 billion in annualized revenue, the market is questioning AI's growth slope

Comparing news, Anthropic's revenue growth is still entering the capital market at an accelerated pace, but controversy surrounding its valuation is also heating up. Bloomberg previously reported that Anthropic had an annualized revenue operating rate of around $65 billion as of the end of July. This figure is still astonishing on the surface, but the focus of market discussions has turned to another level: whether $65 billion means that the growth slope is slowing against the backdrop of some third-party data and optimistic expectations in the AI community pointing at more than $80 billion. The controversy first stemmed from the ARR caliber. ARR, or annual recurring revenue, essentially annualizes the current revenue rate and is not equivalent to audited annual revenue. According to Sacra data, Anthropic's annualized revenue in May was about 47 billion US dollars and rose to 65 billion US dollars in July, but it also warned that revenue from cloud channels such as AWS, Google, and Microsoft may be confirmed in terms of total volume, which will make the scale of revenue seem larger and will also raise the market's focus on gross margin and revenue quality. Optimists still believe that this number is sufficient proof of the strong demand for AI in enterprises. Gavin Baker of Atreides Management believes that Anthropic has had an advantage over OpenAI in terms of token efficiency; PitchBook's Harrison Rolfes points out that even if the model price is higher, if the task success rate is higher, enterprise customers may still accept a higher unit price. In other words, what many people value is not simply the size of API calls, but Claude's ability to pay in the enterprise workflow. Cautists, on the other hand, believe that the market will need to wait for the IPO prospectus to be verified. Simon Willison previously pointed out that run-rate revenue usually comes from short-term annualization of income and cannot be directly regarded as full-year accounting revenue. ThinkFast's Ken Koo also warned that what really matters will be the audit revenue, gross profit margin, customer concentration, computing power procurement obligations, and cash flow in the S-1 file. Ed Zitron more directly questioned that AI companies' ARR may be affected by fluctuations in prepaid tokens, cloud channel revenue, and short-term usage. This discussion isn't just about Anthropic's valuation. Steve Eisman previously referred to OpenAI and Anthropic as key risk points in AI transactions because capital expenses and cloud revenue expectations of tech giants such as Microsoft, Amazon, Google, and Oracle are increasingly tied to leading AI labs.

4d ago

[Comparative Daily News Picks] Yushu Technology will be listed on the Science and Technology Innovation Board on August 19; Anthropic's annualized revenue exceeded 65 billion US dollars before the IPO; Ethereum developers plan to upgrade Hegotá's priority promotion of private transaction proposals in 2027, and FOCIL has confirmed inclusion; the US-Iran situation has added another variable, and the yield on 30-year US bonds hit a new high in 19 years

Daily AI · Crypto · Macro · Market News, Bitpush helps you set priorities ↓ AI · News [Yushu Technology will be listed on the Science and Technology Innovation Board on August 19]. Comparing news, Yushu Technology announced that the company's shares will be listed on the Science and Technology Innovation Board of the Shanghai Stock Exchange on August 19, 2026. [Anthropic's annualized revenue surpassed 65 billion US dollars before the IPO] In comparison, according to people familiar with the matter, Anthropic's current performance means that the company's annualized revenue is expected to exceed 65 billion US dollars, an increase of more than seven times from the level at the end of last year. As of the end of July, Anthropic's annual recurring revenue (ARR) had reached $65 billion, according to people familiar with the matter. One of the people familiar with the matter said that Anthropic shared this data when regularly updating investors on the company's situation. The sharp acceleration in revenue has further strengthened Anthropic's confidence in advancing its listing plan. Both Anthropic and OpenAI have secretly submitted documents related to the listing. Anthropic is expected to land on Wall Street as soon as this fall, possibly earlier than OpenAI. [OpenAI Super Data Center officially launched, Nvidia covered up to 105 billion US dollars] Comparing news, OpenAI's Ohio Super Data Center officially signed a contract. This project was previously revealed. At the time, OpenAI was still discussing a long-term lease with SB Energy, and Nvidia was only considering providing a guarantee. Now that the first 4.25 GW has been officially launched, Nvidia can continue to lock in the remaining 3.75 GW. Previously, the two sides discussed guarantees of up to 250 billion US dollars, and in the end, the initial liability was limited to 105 billion US dollars. This isn't money given directly to OpenAI. Only if OpenAI goes bankrupt or doesn't pay rent, and there is still a gap after the project is re-leased or sold, will Nvidia need to make up the difference. After that, OpenAI will also have to pay back the money actually advanced by Nvidia. Nvidia will also invest $1.5 billion in developer SB Energy, and the park will mainly use Nvidia's AI computing power. Hwang In-hoon estimates that each generation of systems deployed here may correspond to about 1.5 million GPUs and 150 billion to 200 billion US dollars in revenue. Until now, outsiders have been questioning that this model is circular financing: Nvidia backs up the customer's infrastructure, and the customer then uses the money to buy Nvidia chips. Hwang In-hoon also specifically responded this time, stressing that Nvidia only bears specific rent, electricity, and asset residual value risks; it is not responsible for the entire project on behalf of OpenAI. Crypto · Market [Ethereum developers plan to prioritize private transaction proposals in the 2027 Hegotá upgrade, FOCIL has confirmed inclusion] In comparison, Ethereum Foundation researcher Toni Wahrstätter said that the protocol architecture team hopes to prioritize Frame Transactions (EIP-8141) and FOCIL (EIP-7805) in the Hegotá upgrade planned for 2027. Frame Transactions can collaborate with Keyed Nonces and Recent Roots (EIP-8272) and Transaction Assertions (EIP-7906) to enable the privacy pool to pay transaction fees and let the wallet set execution conditions after transaction submission. FOCIL can provide agreement layer inclusion guarantees for eligible transactions. Currently, FOCIL is the only proposal that Hegotá has confirmed inclusion, and the Frame Transactions related scheme is one of 66 proposals currently being evaluated. Other candidate solutions include transaction pricing, status growth, block access lists, and optional zkEVM certification for the main network. Vitalik Buterin previously proposed improving Ethereum's privacy, resisting quantum security, and reducing reliance on second-layer networks. Hegotá will follow Glamsterdam, and the developers plan to complete Glamsterdam by the end of 2026. [CleanSpark, BitFufu, and Canan Technology's Bitcoin production in July fell by about 5%, 10%, and 28%, respectively] In comparison news, according to The Block, Bitcoin mining companies CleanSpark, BitF...

4d agoBitpushNews#Compare Daily Picks

Morgan Stanley is optimistic about Amazon's AI growth potential: AWS may become a trillion-dollar business

Comparing news, Morgan Stanley said that if the growth of the Amazon (Amazon) cloud computing business of AWS accelerates further, the company's stock price may reach $500 by the end of 2027, which is close to double the current level. Morgan Stanley's optimistic scenario is based on how AWS is expected to grow into a $1 trillion business with annual revenue in the future. The main drivers include growing demand for artificial intelligence computing power and data center expansion. As competition for AI infrastructure heats up, Amazon is drastically increasing AI-related investment. The company recently raised its AI capital expenditure forecast for 2026 to $220 billion, focusing on expanding cloud computing infrastructure, AI computing power, and data center capabilities. Morgan Stanley believes that AWS will continue to play a central role in the AI wave, and as enterprises accelerate their adoption of generative AI services, demand for cloud computing and inference computing power may further drive AWS revenue growth. However, Morgan Stanley currently maintains a benchmark price target of $335 for Amazon, which is about 28% higher than the current stock price. The $500 target is a more optimistic scenario, depending on whether AWS can achieve faster growth and fully seize AI infrastructure market opportunities.

5d ago

From crypto mining farms to AI clouds: Why does a16z say the “new cloud” burns money as it grows?

Source: a16z New Media Author: Moses Sternstein, a16z Original title: Charts of the Week: Head In The Neoclouds Editor's Note: In the context of generative AI driving a new round of computing power investment, market discussions on AI infrastructure are shifting from “whether there are enough GPUs” to “who can provide computing power in a sustainable way”. When model training, inference requirements, and data center expansion became consensus, a lower-level question began to emerge: Can the rapid increase in computing power demand actually translate into stable profits and cash flow? In “Charts of the Week” published by a16z New Media, author Moses Sternstein moved in from new cloud companies such as CoreWeave, Nebius, and Applied Digital to discuss the growth, valuation, and profit conflicts of the AI computing power market, and further extended to horizontal SaaS, model routing, and cutting-edge lab talent competition. In this article, instead of simply judging whether AI demand is strong, the author breaks down current AI transactions into a set of lower level structural issues: how existing infrastructure is being repriced, why revenue growth is not simultaneously improving market expectations, and why the AI industry's competitive focus is shifting from simple expansion to efficiency and return. The first is the rediscovery of the value of infrastructure. In the past, land along railway lines, gas pipelines, and cable television networks all served specific industries and were later transformed into telecommunications and internet infrastructure. Today, a similar revaluation of assets happened again. Originally serving cryptocurrency mining, some new cloud companies already have operating experience with electricity, computer rooms, cooling systems, and high-density computing; after the outbreak of AI demand, these capabilities were quickly transformed into scarce computing power supplies. The point is that AI infrastructure competition doesn't start entirely from scratch; early advantages often come from a recombination of old assets, energy resources, and engineering capabilities. Second, high revenue growth and profit uncertainty coexist. The early revenue growth rate of new cloud companies such as CoreWeave once surpassed the initial stages of cloud giants such as AWS, but the capital market did not receive the same level of recognition. The reason is that the new cloud is not a typical asset-light software business. GPU procurement, power access, data center construction, chip depreciation, and debt interest will rise simultaneously with scale, or even faster than revenue. This means that revenue expansion can only prove that AI computing power is in high demand, but it cannot automatically prove that the business model has a sufficiently high return on capital. What the market is really waiting for is whether these companies can turn orders and revenue into sustainable free cash flow. Third, the value of software is being re-differentiated according to the impact of AI. In the past, the market feared that generative AI would generally weaken SaaS companies' moats, but Atlassian's performance suggests AI could also be a tool to increase customer spend and product stickiness. At the same time, cybersecurity and observability software continues to receive valuation premiums as AI expands potential risks and increases companies' reliance on proven solutions. This means that the so-called “end of SaaS” will not happen evenly. Whether AI is an alternative product, lower prices, or expand demand, is becoming the new standard for software valuation differentiation. Fourth, AI applications are shifting from “stacking tokens” to optimizing tokens. In the past, companies often preferred to directly call the most capable models or give engineering teams a budget to test on their own; now, companies such as Databricks have begun to use intelligent routing to match models with different prices and performance according to the difficulty of the task to reduce costs while maintaining results. A decrease in the unit price of tokens does not necessarily mean a contraction in total AI spending: as unit costs decrease and application scenarios increase, total token consumption and overall market size may continue to rise. Efficiency and demand are not mutually exclusive, but may form a mutually reinforcing cycle. If I were to reduce this article to one judgment, it would be: AI infrastructure has proven itself to generate rapid growth, but the next phase of success or failure will depend on whether the company can transform growth into greater capital efficiency. In this sense, the topic discussed in this article is not only whether CoreWeave can become the next generation of cloud giants, but whether the entire AI industry can move from expanding computing power to sustainable commercial returns...

5d ago22#a16z

Stock issuance expanded to $20 billion, and Intel institutional demand is said to exceed $100 billion

Comparing news, Citrini analyst Jukan said in an article on the X platform that the GF Securities Overseas Electronics Newsletter reaffirmed Intel's purchase rating and target price of $136, and believed that the stock offering would send a positive signal. The report predicts that Intel's foundry business will achieve break-even in the fourth quarter of 2027, and the profit margin leverage effect will be fully reflected in 2028. Yield and external customer expansion, particularly Apple-related developments, are progressing steadily, and EMIB's customer base is also expanding, including Google and AWS. Intel has received support from the substrate vendor Unimicron and will manufacture silicon capacitors in-house. The report raised Intel's 2026 and 2027 earnings per share expectations by 3% and 1%, respectively, and the target price remained unchanged at $136 after considering the effects of dilution. Intel expanded its stock offering from the initial plan of $15 billion to $20 billion, and institutional demand is said to have surpassed $100 billion. The issue price is $95, and all over-allotment rights have been exercised. CEO Lip-Bu Tan and his family subscribed for approximately $12 million at the issue price. The report believes this reflects management's confidence in the company and may support FY2027 capital expenditure. According to the report, Intel reiterated that the foundry business will achieve break-even by the end of 2027; if postponed to 2028, the main reason will be increased demand for new investment. The report maintains its previous judgment. The yield of 18A is expected to be around 80% in the second quarter of 2026, and CWF has entered a phase of climbing capacity. Cooperation with external customers continues to advance, and Apple's large-scale mass production of the 14A is particularly prominent. The report raised Intel's FY2027 and FY2028 back-end business revenue estimates to $1.1 billion and $7 billion, respectively, due to increased visibility into AWS Trainium3's adoption of EMIB-T in 2027, and Google's Humufish and Triggerfish will enter a phase of production expansion from the second half of 2027 to 2028. The report also predicts that AWS and Microsoft's ASIC products may adopt EMIB in 2028.

8d ago

Why did NeoCloud rise more sharply than Nvidia in this round of technology stock rebound?

Author: Vibrant BlockBeats Original title: Why did NeoCloud increase the most in this round of rebound in US technology stocks? One of the strongest directions in this round of US tech stock rebound came from NeoCloud: CoreWeave, Nebius, and some AI infrastructure companies with power and data center resources. Logically, the capital is pricing an AI infrastructure equity certificate with multiple leverage: computing power production capacity that has been locked in a contract and can be delivered quickly. Once AI demand improves, NeoCloud's revenue expectations, financing capacity, and shareholder equity value are likely to rise at the same time. This makes it highly resilient during the rebound phase of technology stocks; electricity, data centers, financing, and valuation flexibility together form this level of leverage. The AI bottleneck is changing. What was most scarce in the early days was GPUs, followed by HBM and high-speed networks; today, what customers really lack is a complete set of capabilities to go online: get a GPU, have enough power, complete computer room construction, network connectivity, and be able to deliver large-scale clusters within a few months. NeoCloud is stuck in this gap. The funds were purchased by NeoCloud, a “powered computing power factory,” usually including GPU clusters, networks, liquid cooling, data centers, power access, and operation and maintenance services. The customer purchased a block of large-scale computing power capacity that can directly run AI training and inference. This is important. GPUs can be purchased, but power capacity, land, substations, data center licenses, and network access cannot be replicated in the short term. Large cloud vendors have capital and customers, and are also bound by the construction cycle; some AI companies want to preserve more flexibility and are unwilling to put all of their needs on a single hyperscaler. As a result, NeoCloud, which has ready-made electricity and rapid deployment capabilities, became an “accelerator” for investment in AI infrastructure. The market is willing to value them higher, and the core is that these resources have two characteristics: · Scarce: limited available electricity and deliverable data center capacity; · Contractable: customers are willing to sign multi-year capacity contracts with minimum commitments. When scarce resources can be locked in by long-term contracts, the market will reinterpret it from ordinary IT service revenue as a cash-flow asset with infrastructure attributes. Financial reports have changed the market's view on the business model. Previously, the market's main question about NeoCloud was very direct: buying GPUs and building data centers required huge amounts of capex. Will the company fall into a cycle of “continuous financing and continuous burning of money”? The answers given in recent financial reports were positive. CoreWeave Q2's revenue reached $2,575 billion, disclosing a backlog (signed but unconfirmed expected revenue) of approximately $104 billion; Nebius' AI Cloud ARR (annualized recurring revenue) reached $3 billion, and disclosed a number of large long-term contracts. The market focuses on single-quarter revenue, and more on the complete commercial loop that appears behind these numbers: AI customers sign long-term capacity contracts → some customers provide advance payments or minimum payment commitments → companies can more easily obtain debt and equipment financing → add GPUs, computer rooms, and power capacity online → revenue and EBITDA (profit before interest, tax, depreciation and amortization) increase → continued increase in financing capacity and expansion capacity. This has gradually moved NeoCloud's narrative from “high-capex GPU renters” to “AI that supports expansion with orders” “Infrastructure operators”. As long as orders, financing, and delivery can continue to be linked, growth will have a clear flywheel character. Why isn't funding prioritizing storage and the three major clouds? The choice of funding reflects poor expectations in different areas. Storage leaders are benefiting from AI demand, and products such as HBM and DRAM are still very popular. However, the market has begun to worry about rising supply, high prices, peaking profit margins, and whether upbeat expectations in the early period have been fully reflected in stock prices. The financial report is strong. If the forward guidance does not continue to be revised, the stock price will easily be under pressure. The challenge for storage companies is their cyclical nature. The market deals with prices, shipments, and gross margin paths for the next few quarters; when supply is likely to catch up with demand and average selling prices may fall, it is difficult for strong current performance to continue to drive valuation expansion. HBM/DRAM, NAND/SSD, and HDD are also in different sub-cycles, and the stock price performance of all storage companies cannot be attributed to the same reason. Three major clouds — Microsoft Azure, Amazon...

9d agoburnking#AI #Arithmetic power #US stocks #financing

The “thought process” of closed source AI has been stripped away. How did the most valuable moat collapse?

Author: Claude, Shenchao TechFlow Original title: The latest paper sparks discussion: The AI “deep thinking” process can be distilled for free, and the most valuable training asset of closed source manufacturers is being emptied Deep Wave Guide: Every time you ask a question to AI, it will first “think deeply” and then open up in the background. This thought process that no one can see is the moat at the bottom of OpenAI and Anthropic's pressure box. Now, a group of researchers has revealed a way to take this thought out completely. Also stripped out were the credit card numbers, passwords, and email addresses posted by users. This isn't a security paper far from you; it's a sign that the way you talk to AI in the future may change. On August 10, a paper was submitted to arXiv, and the next day, the project website stolen-thoughts.com was launched, showing “thought records” taken from multiple closed source models one by one, which reached 500 points on Hacker News. Project leader Alexander Panfilov wrote on X: “We've found a way to extract the hidden inference of cutting-edge models by exploiting bugs in all cutting-edge AI companies' APIs.” In other words: you think only AI knows what it's thinking; in fact, someone can unfold it. What you put in AI may be leaking along with its “thoughts”. The researchers scanned about 7,000 AI assistant session records that were publicly shared online, unraveled the encrypted “thought process” one by one, and then discovered some things that should not have appeared. Panfilov tweeted: “We initially scanned around 7,000 public conversations and found 62 unique API keys, 33 email addresses, 33 passwords, and other sensitive information.” Even more glaring are the details. In the “thought” of a flight booking task lies the full name, email address, passport number, date of birth, and credit card number with a security code. Keys for platforms such as Anthropic, AWS, and GitHub also appeared in the case. In other words, the credentials you put in for AI to help you do your work will be saved along with its thought process, and then taken away by others. “If you've ever shared Claude Code or Codex sessions with encrypted inference blocks online, they can all be decoded and reveal your personal data.” Panfilov wrote. The most direct reminder to regular users is: stop posting secrets to AI, even if it says “I won't spread it.” The manufacturer first said “it's OK,” then secretly fixed it. This incident did not happen suddenly. Matthew Green, a professor of cryptography at Johns Hopkins University, reported a similar vulnerability to the manufacturer in May of this year, and the response received at the time was “no security impact was seen.” By the time the Panfilov team officially revealed it, the manufacturer's attitude changed. “We have since gone through the responsible disclosure process. The vendor has fixed a number of issues caused by this vulnerability, which, as far as I know, is continuing.” Panfilov said. The paper also confirmed that after disclosure, the researchers were no longer able to reproduce the same attack. The problem is: the vulnerability existed for a few months, and users didn't know about it. The fix will not actually be implemented until it is revealed and discussed. It's not just one company's fault; it's the first time that the industry's “encryption is security” assumption has been publicly debunked. For readers, what's really worth remembering is the saying: the AI company you trust probably didn't tell you all the risks. The model you use is probably no longer so “exclusive” and has changed over a longer period of time at the industrial level. Reasoning ability is the foundation of OpenAI and Anthropic's pricing, and it is also the part they are least willing to reveal. Once this thought can be extracted in batches, competitors can feed the “ideas” of the strongest models to their own models to learn at a very low cost. The paper also mentions a preliminary observation that has not been peer-reviewed: using the “thinking” of a small number of the strongest models to guide another model will clearly drive the latter's answers in the direction of the former. What does this mean? The closed source model moat originally meant “you can't buy my brain with money.” There is now a crack in this wall. There's nothing bad about users in the short term: stronger competitors may come up faster, and prices may be knocked down. But the cost is that you can no longer tell if a model is really smart or has copied someone else's idea. Who exactly does the “thought you pay but can't see” belong to,...

10d agoburnking#AI #Anthropic #OpenAI

AI is being used to create a virus for the first time. The US Senate warns AI giants: forcibly intervene without suspension

Author: Shenchao TechFlow Original title: US Senate warns AI Big Three: Immediate suspension of AI development, otherwise Congress will force intervention in Shenzhen Guide: US Senator Bernie Sanders (Bernie Sanders) wrote to OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Meta CEO Mark Zuckerberg on August 10 requesting an immediate suspension of AI development. Citing incidents such as the AI model getting out of control and invading external systems and AI being used to create a novel virus for the first time, Sanders pointed out that the three companies' previous security promises had been triggered, and “that moment has arrived.” He warned that if companies do not act on their own, the Senate will step in on their behalf. On the same day, 29 Democratic members of the House of Representatives also sent a joint letter to OpenAI and Anthropic requesting an explanation of the loss of control of smart devices. AI regulation is being upgraded from a technology issue to a bipartisan political agenda. The US Senate's patience with Silicon Valley's AI giants is running out. According to an exclusive report by Axios on August 10, Vermont Independent Senator Bernie Sanders sent letters to OpenAI CEO Sam Altman, Anthropic CEO Dario Amodei, and Meta CEO Mark Zuckerberg on the same day. The first batch of letters was made public to Axios. In his letter, Sanders bluntly asked the three CEOs to deliver on their respective companies' security promises and immediately suspend AI development. “For the benefit of humanity, please keep the promises you have made. Pause AI development. It is not too late to act; disaster can still be avoided. Stop making machines that humans can't control.” In his letter, Sanders wrote, “I want to make it clear: if you don't take appropriate action now, my colleagues in the US Senate and I will act.” The AI model got out of control and was first used to create a virus. Sanders listed the three major triggering events. Sanders listed three recent key events in the letter as a basis for applying pressure. First, around August 6, a research team used AI for the first time to design a complete functional viral genome never seen in nature. Sanders warned that such technology “could kill tens of millions of people if it falls into the wrong hands.” Second, in July, OpenAI's Sol model “got out of control” during cybersecurity tests, escaped from an isolated sandbox environment to access the Internet, and eventually hacked into the technology company Hugging Face's system. OpenAI called it an “unprecedented cybersecurity incident” at the time. Sanders pointed out that the act “clearly violates federal law.” Third, after completing an internal review, Anthropic and Meta also reported similar incidents of models falling out of control. What the three events have in common is that the AI model broke through the limitations of the operating environment without instructions. A misconfiguration by cybersecurity company Accidental has been cited as the cause of some of the incidents. The hacked company called this AI hack an “unprecedented incident” and required “unprecedented countermeasures.” Yoshua Bengio (Yoshua Bengio), the world's most cited living scientist, said these events “should be a wake-up call.” Sanders clearly agreed in the letter. The three companies' safety promises have been named one by one, and the “critical moment” has triggered the Sanders letter's core strategy to “attack the shield with the spear.” He quoted each of the three companies' previously disclosed security promises, pointing out that the trigger conditions for these promises have been met. In 2023, Anthropic stated that the company would “commit to suspending scaling and/or delaying deployment of new models when the expansion capacity exceeds our ability to comply with security procedures.” In 2025, Meta stated: “If a cutting-edge AI is assessed as having reached a critical risk threshold and cannot be effectively mitigated, we will stop developing it.” In the same year, OpenAI said that if AI capabilities reach a “critical” threshold, the company will “stop further development” until strong security measures are in place. Sanders wrote, “That moment has come. AI capabilities have reached a critical threshold.” At the same time, he quoted the statement of John Ratcliffe (John Ratcliffe), director of the US Central Intelligence Agency (CIA) as circumstantial evidence. Ratcliffe is 6 this year...

11d agoburnking#AI #Anthropic #OpenAI #custodial